Formal Techniques for Synchronized Fault-Tolerant Systems

Abstract

We present the formal verification of synchronizing aspects of the Reliable Computing Platform (RCP), a fault-tolerant computing system for digital flight control applications. The RCP uses NMR-style redundancy to mask faults and internal majority voting to purge the effects of transient faults. The system design has been formally specified and verified using the EHDM verification system. Our formalization is based on an extended state machine model incorporating snapshots of local processors clocks.